Reliability problems with an array of 91 high-power LEDs

Question

I have designed and printed a 4-layer PCB that accommodates 91 infrared LEDs in a 7x13 rectangular layout. This will be used as a backlight for a machine vision project. I am having a problem where individual LEDs are burning out or perhaps becoming disconnected from the circuit in some way. I suspect the heat dissipation may be the cause of the problem.

Image

PCB Layout

Each row of 7 LEDs (green LED text) is wired in series. The 12V supply (VCC powerplane) connects to the first LED. The next 6 are wired in series. Finally a current-limiting resistor (green R text) connects the last LED to the ground plane.

Specifications:

VCC plane: 12V, 2A supply
LED: TSHG6200. 100mA maximum rated current.
Current limiting resistor: 20 ohms
Solder: Thermoflow Sn60/PB40
Total estimated power dissipation: 12V * 0.1A per row * 13 rows = 15.6W.
Size of array: 13 rows of 7 LEDs, approximately 7cm x 6cm

Measurements

With a 12V power supply, there is about 1.45V over each LED, and about 2.0V over the current limiting resistor, meaning a current of 100mA. Because this is right at the maximum allowable current, I put a big high-power potentiometer between the power supply and the VCC plane, and used this to regulate the input voltage to be slightly lower (11.5V or so). This gets the current safely below the maximum allowable amount.

I am also using a Darlington pair to control the backlight with an Arduino. The backlight is on almost all the time, and occasionally is pulsed off for about 30ms. I don't think this is relevant to the problem but can provide more details if necessary.

Problem

After about 10-30 minutes of use, one or more of the rows of LEDs will go out. If I measure the voltage across each LED in the broken row, most LEDs are at about 0.8V and one has about 8.0V across it. No current is flowing. Sometimes resoldering the pins or tapping the LED fixes this. Sometimes it has to be replaced. In any case I only get another 10-30 minutes of use before another one goes out.

Another observation is that the whole back side of the board is kind of sticky. You can see this in the picture above. I wonder if it is getting too hot and the solder is becoming compromised (perhaps exuding flux??).

Question

What should I try to improve the reliability? I've already tried running it at a lower voltage to get the current safely below the rated maximum. I wonder if I need to use a different kind of solder? Or some kind of heat sink? The LEDs get hot to the touch but not unbearably so.

Edit, after trying suggestions

Thanks everyone for the tips! I did something quite simple -- pointed a computer fan to blow air across the array -- and it worked fantastically! I guess this is really obvious to many of you but I was surprised at how enormous the difference was.

Without fan:

25mA per row -> 39C
33mA per row -> 41C
40mA per row -> 48C
55mA per row -> 52C

So we get into the "danger zone" of temperature well before reaching the maximum current per LED.

With fan:

35mA per row -> 26C
60mA per row -> 30C
90mA per row -> 34C

I ran it at 90mA per row and 34C for over an hour with no problems. Great!

Where did you purchase the LEDs from? eBay? Did you purchase them from a relaible supplier? — Gilad, Apr 07 '16 at 17:58
There are SMD high power LEDs that accept multiple lenses (for wide beam, diffusion, etc), they can be mounted on metal-core PCBs improving reliability a lot. Higher power also means less LEDs which will make it easier for your assembly. — Wesley Lee, Apr 07 '16 at 18:50
@WesleyLee what is a metal-core PCB, is that the same as the 4-layer PCB that I am using with a ground and power plane? — cxrodgers, Apr 07 '16 at 19:35
its a PCB with an alu alloy instead of fiber-glass, for heat intensive applications: http://www.electropages.com/wp-content/uploads/article_images/large/12464.jpg — Wesley Lee, Apr 07 '16 at 20:16
Where I'm from its mostly limited to single sided designs but for your application one side would suffice. — Wesley Lee, Apr 07 '16 at 20:17
As others have said, the main problem is undoubtedly thermal dissipation. One additional issue is that the pads look a bit small. They might be fine for flow soldering, but for hand soldered stuff a larger pad size would help, and would probably also improve thermal dissipation. You have plenty of space for bigger pads- use it. — JavaLatte, Apr 07 '16 at 20:35
Even if just pointing a fan at your current board is working as an interim solution, if you're going to be making more of these I'd suggest spacing out the LEDs a bit more. The way you currently have them packed together, there's not much space between them for airflow, so even the fan won't be as effective as it could be. — Ilmari Karonen, Apr 08 '16 at 10:25
May I suggest that you'd probably be much better off using something like this: http://www.aliexpress.com/item/1W-3W-5W-10W-20W-30W-50W-100W-200W-IR-850nm-Infrared-High-Power-LED-Lamp/32642955843.html — Jules, Apr 08 '16 at 14:27
To add to the comments of others: lots of surface-mount LEDs include a heatsink pad in addition to the cathode and anode pads. So you can design thermal vias into your PCB and mount a heatsink on the other side. — John Walthour, Apr 08 '16 at 15:01
Only one resistor? If so, expect significant variation in how much current the different LEDs get. — user6030, Apr 09 '16 at 00:09
@Jules I'd be careful of where you source those things. See https://www.youtube.com/watch?v=NjKgPLeJ79Q - might be some QC issues there. — Bob, Apr 09 '16 at 10:11
@user6030: There is one resistor per _series chain_ of LEDs as is standard practice. Each LED in the chain will have the same current. — Transistor, Apr 09 '16 at 13:04
how about using the 'legs' of the led as a heat sink/dissipator?...leaving a gap of say 5 or 6 millimetre bottom of the led and the board. — , Apr 09 '16 at 11:56
@Al-Nizam That would be fine, but you need to make sure you have enough air flowing past those legs in order to remove the heat without excessive temperature rise. Not a trivial proposition. — Dave Tweed, Apr 09 '16 at 13:45
By the way, the sticky stuff on the back looks to me like flux left after soldering with flux-core solder. The fastidious engineer removes this after soldering using perhaps an acid brush with bristles cut short, isopropyl alcohol, and a suitable absorbent sheet (Kim-wipes are my favorite). I personally use a 50-50 mix of 99% Isopropyl + Acetone which works a bit quicker on rosin fluxes. The acetone will attack certain surfaces though, so some care is required. I've only had trouble with it on plastic housings and stripes on cheap resistors; all the other components seem to be impervious. — scanny, May 08 '16 at 21:04

Transistor · Answer 1 · 2016-05-08T13:36:12.950

30

From the datasheet:

Figure 1. Absolute maximum forward current.

and further on:

Figure 2. Derating LED current at increased temperature.

Current and temperature are your problems. You are running at absolute max current with no wriggle-room and you are allowing the temperature to rise. At 60° ambient the max current allowed falls off dramatically.

edited May 08 '16 at 13:36

answered Apr 07 '16 at 18:41

Transistor

168,990
12
186
385

1

I accepted the other answer because of some of the additional info in it but I just wanted to say this graph was really helpful for me to understand the problem – cxrodgers Apr 08 '16 at 01:10
2

It is worth mentioning that the graph does not say once the LED has reached 60 degrees C, the current must be derated. It's not. It is saying if the ambient temperature of the environment the LED is shedding heat into is 60 degrees, it must be derated. Now, this is more or less the case as each LED is surrounded by a bunch of other hot LEDs, which will do a great job of simulating a hot ambient environment. It's a distinction worth making, however. – metacollin May 08 '16 at 13:00
Good point. I've added the word "ambient" into the last sentence to clarify. – Transistor May 08 '16 at 13:37

WhatRoughBeast · Accepted Answer · 2016-04-07T18:52:36.210

23

You have already hit on the answer: your LEDs are getting hot. 15 watts may not sound like much, but it's building up and killing your LEDs. I suggest you get a thermistor and attach it to the center of the board, then monitor the temperature as the system operates. Even better, attach it to the body of one of the LEDs.

Because you're using this as a backlight, don't use narrow-beam LEDs. Use relatively wide beam units, and space them apart so air can flow through. If you can find a source of, let's say, 35 degree LEDs, install only every other one in a checkerboard pattern, soldering jumper as necessary. You'll only get half the total brightness, but that's barely perceptible, and the improved airflow should be a big help. You may also need to provide a fan with some ducting to keep the air flow through the array adequate.

And always include a temperature monitor. While not directly applicable, this YouTube video shows the principles of cooling. In your case, since you've got a forest of vertically standing LEDs, it is important not to let the LEDs touch each other, since this will block the flow of air.

edited Apr 07 '16 at 18:52

answered Apr 07 '16 at 18:13

WhatRoughBeast

59,978
2
37
97

I will try to install a thermistor as you suggest. I wonder if a fan or a heatsink would be easier for me to get working. I'll check out the video. And point taken about the greater spacing. – cxrodgers Apr 07 '16 at 19:39
4

For what you're doing, heatsinks just don't work. There's no way to attach them. SMD LEDs can be made to work, though. – WhatRoughBeast Apr 07 '16 at 20:08
1

Get it working again, and point a good-powered fan at it. See what that does to the operating time. I concur, 15W doesn't sound like much, but that's enough to get the tip of a soldering iron to 650°F in 5 minutes. – rdtsc Apr 07 '16 at 23:19
1

Problem is, if it's a backlight, it will probably be sandwiched behind a display of some sort, so only cross ventilation will work. Airflow will be much less than if you're just blowing a fan across the array. – WhatRoughBeast Apr 07 '16 at 23:26

score 1 · Answer 3 · answered Apr 08 '16 at 10:00

A quick google for "LED heatsink" will find you a lot of suitable solutions for cooling down your burning-out LEDs. My initial recommendation though would be to turn the top of the PCB into a heatsink layer and use thermal adhesive to stick the LEDs to it. I'm afraid that's more soldering, but hey ho.

For more info, I've found that Wikipedia has a link to various LED cooling tech which may be of interest.

score 0 · Answer 4 · edited Jan 30 '20 at 18:39

I have used arrays of high powered LEDs as an inexpensive way to UV cure conformal coating and an inexpensive light source for solar cell testing. Initially, I attached the LEDs to a 1/4" thick piece of aluminum sheet to remove the heat, and the aluminum and LEDs got very hot. One of the LEDs burned out and had to be replaced. I initially used wet paper towels to cool the aluminum sheet, but this was a bandaid solution. My solution for cooling was to place a copper heat pipe CPU cooler on the backside of the aluminum sheet. This worked very well; the temperature of the aluminum sheet was maintained 30C and the LEDs stopped burning out. These coolers are inexpensive, effective and easy to install.